Skip to content
Longterm Wiki
Index
Division·6J1yGK2zYY·Record

Apollo Evals

Verdictunverifiable85%
1 check · 4/29/2026

1 → unverifiable

Our claim

entire record
Parent Org
Apollo Research
Name
Apollo Evals
Division Type
team
Status
active
Notes
AI safety evaluations focused on detecting deceptive and scheming behaviors in frontier models. Published influential research on in-context scheming in 2024.

Source evidence

1 src · 1 check
unverifiable85%Haiku 4.5 · 4/24/2026

NoteThe claim describes a division/team called 'Apollo Evals' as a distinct entity within what appears to be a larger organization. However, the source is about Apollo Research itself (the parent organization), not about its internal divisions or teams. The source does not mention 'Apollo Evals' as a named division or team. While Apollo Research clearly has evaluation work (Model Evaluations is listed as one of their research areas), the source does not confirm the existence of a formally named division called 'Apollo Evals' or provide information about its type or status. This is a subject-identity mismatch: the claim is about a sub-unit (Apollo Evals), while the source discusses the parent organization (Apollo Research). Without explicit mention of 'Apollo Evals' as a division name in the source, the record cannot be verified.

Case № 6J1yGK2zYYFiled 4/29/2026Confidence 85%